Machine Learning Intern

About the role

Build reproducible machine learning experiments for document and automation problems. Start with a useful baseline, measure what changes and make limitations clear.

What you’ll work on

• Translate a focused problem into a baseline, an evaluation dataset and explicit success criteria.
• Prepare data and build training or inference pipelines that can be reproduced.
• Compare model approaches against simpler methods, with checks for leakage and class imbalance.
• Inspect incorrect outputs and communicate quality, latency and resource trade-offs.
• Package a useful experiment behind a small interface or API when appropriate to the project.

Core skills

• Python, NumPy, pandas and scikit-learn for practical modelling and evaluation.
• A foundation in probability, linear algebra, supervised learning and overfitting.
• Working knowledge of PyTorch or TensorFlow for model training and inference.
• Train, validation and test design; task-appropriate metrics and reproducible experiments.
• Readable Python, Git and the ability to explain a baseline, a result and its limitations.

Useful additional experience

Helpful, but not required.

• Text embeddings, information retrieval or transformer models.
• Evaluation of AI-assisted extraction or retrieval workflows, including unsupported outputs.
• Experiment tracking, model serving or inference optimisation.

Where this work connects

Arvi belongs within the Xind ecosystem and supports document intelligence and PDF-to-HTML workflows. This track would explore document understanding and evaluate possible approaches before any product integration.

Show us your work

Share a model experiment or ML application with code, evaluation data description and results against a baseline. Explain what you changed, how you tested it and where the approach still fails.

Tell us what you contributed, what you learned and what you would improve. Personal projects and coursework are welcome.